doubling resolution strategy # 2

Last year ( Wed Apr 12, 2000) I suggested a way to arbitrarily
increase the effective resolution of an encoder by using pulleys.

Art Eckstein successfully used it in "doubling" mode to double
the resolution of the position feedback of his lathe:
http://members.surfsouth.com/~axtein/dro/dro1/index.html

Here today, I present another way to double (or more) the
resolution of the position feedback:

--- Use two encoders instead of one! ---

This works for both linear encoders and rotary encoders giving out
quadrature signals.

The first encoder gives a quadrature TTL output. Mount the
second in any position, such that its output is displaced by one-half
the given resolution of the first encoder. This can easily be done
for linear encoders by mounting them together, and adjusting
their difference apart by slight sanding, or using a shim. The
actual distance they are apart doesn't matter, just that their
cyclic quad signals are offset slightly. You can use either a
dial gauge, or a 2 channel oscilloscope to get it set up right.
(All this is theory BTW, I haven't done it yet in practice...).

For rotary encoders, procure another read-head, and position it
anywhere, again giving the required phase shift.

http://groups.yahoo.com/group/CAD_CAM_EDM_DRO/files/two_encoders_to_double_quadrature_resolution.jpg

Combine the first two quadrature output signals (A and B from encoder #1),
XOR (exclusive or ) them, and do the same with the second two quad
signals (C and D). Any TTL, such as a 74F86 hex xor chip should work.

These two signals (A xor B) and (C xor D) become the new pair of
quadrature signals, for input into your DRO display or DRO PC-ISA input board.

This idea can be extended to more encoders, but the precision of the
scale becomes increasingly critical, and therefore using more than two
encoders is probably best kept for the rotary encoder versions.

As a practical example, the http://www.usdigital.com linear encoders
have a max resolution of 360 CPR (cycles per revolution), or 4*360 = 1440
quadrature / inch - giving a resolution of 0.000 694444 inch per step
(which is right on the limit of being acceptable to me).

Using two of the $30 HEDS heads, can give you half that, 0.000 3472222
inch per step (which I'm much for satisfied with....).

Using a 2048 CPR rotary disk (8192 in quadrature, but 16384 using two
encoder heads, and a 5 TPI screw) gives an incredible 16384*5= 81,920
steps per inch, or 0.000 012 207 inch/step (or even half that if the
encoder is on the motor with a 2:1 pulley reduction drive).

At these resolutions, it will become increasingly desirable for all of us
to invent some kind of super-precise calibration method, in order to
correct for screw (and encoder scale) inaccuraccies in software.

Doug Fortune
http://www.cncKITS.com